Lecture 6-3: Structural Sin and Automated Injustice

Isaiah 10:1-2 — "Woe to those who decree unrighteous decrees, and to the writers who write oppressive decrees; to deprive the needy from justice, and to rob the poor among my people of their rights, that widows may be their plunder, and that they may make the fatherless their prey!" (World English Bible)

Ephesians 6:12 — "For our wrestling is not against flesh and blood, but against the principalities, against the powers, against the world's rulers of the darkness of this age, and against the spiritual forces of wickedness in the heavenly places." (WEB)

3.1 Why "no one intended it" is not the end of the theological question

A common and understandable response to a case like Gender Shades or the Obermeyer healthcare algorithm (lecture 6-2) is to note that the harm arose from unrepresentative data and flawed proxies rather than from anyone's deliberate malice, and to conclude from this that the wrongdoing, whatever it is, is not really sin in the theological sense — sin, on a common intuitive picture, requires a sinner: an agent who knew better, chose otherwise than they should have, and bears guilt for the choice. On that picture, an unintended statistical artifact seems to belong to a different category altogether — an error, a bug, a regrettable accident, but not sin.

This lecture argues that picture of sin, while capturing something true about personal sin, is too narrow to be the whole doctrine, and that Christian theology has never actually confined itself to it. Two resources, one modern and one ancient, are worth engaging directly.

3.2 Structural sin: the liberation-theology contribution

Latin American liberation theology, developing through the 1970s and 1980s (Gustavo Gutiérrez's A Theology of Liberation, 1971, is the foundational text), pressed the church to name a category it argues Scripture assumes but systematic theology had under-developed: social or structural sin — sin embedded not in a discrete act of an individual will but in an institution, a system, or a set of social arrangements that produces unjust outcomes as a matter of ordinary operation, independent of any single actor's malicious intent. A tax system, a land-tenure arrangement, an economic order — each can be structured such that it reliably produces poverty or exploitation for a class of people, and each participant within it can be, individually, no worse than ordinarily selfish, while the system as a whole does real and repeated harm. Liberation theologians argued the church's traditional, almost exclusive focus on individual, confessable sin had made it poorly equipped to name this kind of wrong at all, let alone repent of it or resist it.

The category has not been without controversy — critics, including from within Catholic magisterial responses to liberation theology in the 1980s, worried that "structural sin" language could be used to diffuse personal moral responsibility entirely into "the system," or could be annexed too readily to a particular political-economic analysis (frequently Marxist-influenced in its original context) that a reader need not accept in order to accept the theological category itself. This lecture uses the category in the narrower, more defensible form: not a claim that any specific economic system is uniquely sinful, but the more modest and, on reflection, more obviously true claim that an unjust outcome, produced reliably by a system's ordinary operation, is a real theological wrong requiring a real theological name — a name broader than "someone's individual, culpable choice," because that is often, honestly, not what happened.

An algorithm trained on historically discriminatory data and deployed at scale fits this category with almost no adjustment needed. No loan officer decided to disadvantage a demographic group; the historical data simply carried forward what centuries of a discriminatory housing and lending market had already produced, and the model reproduced it faithfully. That is structural sin in close to its purest contemporary form: harm, reliably produced, embedded in a system's normal function, disconnected from any single culpable decision.

3.3 Principalities and powers: an older resource for the same instinct

Liberation theology's structural-sin language is not, however, the only or even the oldest biblical resource available here, and a reader coming from a more traditional theological framework — one wary of importing twentieth-century political theology wholesale — has an alternative that reaches the same basic conclusion from a different direction. Ephesians 6:12 names "principalities," "powers," "the world's rulers of the darkness of this age," and "spiritual forces of wickedness in the heavenly places" as the real object of Christian struggle, explicitly contrasted with "flesh and blood" — that is, explicitly contrasted with a struggle against individual human opponents. Pauline scholarship on this language (Walter Wink's trilogy on the powers, beginning with Naming the Powers, 1984, is a frequently cited if contested modern treatment) has long argued that "principalities and powers" names something like the spiritual dimension of institutions and systems themselves — not merely personal demonic agents operating apart from human structures, but a spiritual reality that can inhabit, distort, and outlast the individuals who staff an institution at any given moment. On this reading, a legal system, an economic order, or — with no great interpretive strain — a widely deployed algorithmic system can be a vehicle through which "the powers" operate, doing real harm that no single human actor within the system intended or could unilaterally undo.

The value of naming both resources side by side, rather than only reaching for the more recent one, is exactly what this course's non-negotiable commitment to fairness across traditions requires (see the "contested points" discussion elsewhere in this course): a student from a tradition suspicious of liberation theology's political associations still has, in Ephesians 6:12, an old and entirely orthodox textual basis for taking structural, more-than-individual evil seriously. The two resources do not need to be forced into agreement on every point of their broader theological frameworks to converge on the narrower claim this lecture needs: Scripture already has categories for sin and evil that exceed any one person's choice, and a biased algorithm is a plausible contemporary instance of exactly what those categories were meant to describe.

3.4 "Woe to those who decree unrighteous decrees"

Isaiah 10:1-2 was addressed to Judah's political and judicial leadership — those who "decree unrighteous decrees" and "write oppressive decrees," with the effect of depriving "the needy from justice" and making "widows... their plunder" and "the fatherless their prey." The prophet's target is not a corrupt individual verdict in an isolated case; it is policy — decrees, written and issued at the level of a governing class, applied broadly and repeatedly to whoever falls under them, with predictable victims named explicitly (again, the widow and the fatherless, continuing the pattern from 6-2).

Read alongside the argument of this lecture, that description is close to a literal account of what an algorithmic decision system is: a decision rule, written once by a small group of people, applied at scale to a large population, with statistically predictable effects on the people least positioned to contest or even notice it. Isaiah's charge was against decrees inscribed on scrolls and enforced by human officials; the mechanism of oppression this lecture has been describing is decrees inscribed in code and enforced by servers, at a speed and scale the prophet could not have imagined but would very plausibly have recognized in substance. The text does not need to be stretched or allegorized to reach this application — it already names, in its plainest sense, a policy that produces predictable injustice for the vulnerable at scale, which is a serviceable one-sentence definition of what an unaudited biased algorithm does.

3.5 The question turned toward the church

The chapter closes by turning the diagnostic inward. It would be a comfortable but incomplete ending to this chapter to leave "algorithmic bias" as a problem belonging to banks, hospitals, and technology companies. Churches and ministries increasingly use software systems that make or inform decisions about people: donor-management platforms that score or rank giving patterns, background-check and screening services for volunteers and staff, attendance-tracking tools that flag "disengaged" members, even AI-assisted tools for evaluating grant or scholarship applications. Each of these is a smaller-scale instance of exactly the three pathways from lecture 6-1 (training data, proxy variables, feedback loops), and each is exactly as subject to Isaiah 10:1-2's charge and Leviticus 19:15's demand for impartial judgment as any secular institution examined in this chapter. A church that would rightly criticize a bank's unaudited lending algorithm has an obligation, following from everything argued in this chapter, to ask the same audit question of its own administrative systems — not as an act of institutional self-flagellation, but because the biblical justice texts were never written to exempt God's own people from the scrutiny they apply to everyone else.